Sphinx uses a navigation style based after books. Each page links to the next and the one before it. This is all well and good for linear page flow, but it makes for clunky API docs.
We should look into what it takes to get arbitrary html/css content on our pages. Sphinx is output agnostic (compiling reStrucured text into html/pdfs/others) so this is cutting against the grain a bit, but sphinx was also designed to be extensible so this should be possible.